Ideology, Nationalism, and Identity in Basque Regional Elections

Stephen Ansolabehere Maria Socorro Puy Segura

May 14, 2013

Abstract Parliamentary elections to the Basque Autonomous Community have a stable multi- party system that regularly produces long-lived minority and coalition governments. More amazing still, this stable party system arises in the context of a complex social and political setting in which the society cleaves along at least two lines – ideology and nationalism – and in which people have strong identities tied to the Basque language and culture. This paper analyzes voting behavior in parliamentary elections in this region to understand how ideology, nationalism, and identity sustain this party system. We extend the conventional spatial voting model to incorporate identity issues. Our empirical analysis shows that ideology, nationalism (or regional autonomy) and identity strongly predict vote choice, and, surprisingly, that ideology is much more salient in vote choice than nationalism. Interestingly, the analysis suggests that identity politics both polarizes voting and sustains a stable multi-party system.

1 1 Why Basque Parliamentary Elections

The Basque region of Spain holds a distinct fascination. It boasts an ancient language with mysterious roots, one of the earliest recorded representative assemblies in Eu- rope, and a long history of occupations and resistance. Today, this region of just over 2 million people has become a model for regional government in Europe. The Basque Autonomous Community has emerged as one of the most economically successful re- gions of Spain with a relatively stable political system. The BAC is closely eyed by regions of other countries where local political leaders seek either greater autonomy or even independence.1 For Political Science, Basque Parliamentary elections present an interesting, yet puzzling, case. Basque regional elections are an excellent, clear case of multi-party politics in a multi-dimensional setting. Typically three to four main parties divide 90 percent of the votes and seats, and no party has ever won a majority of seats in the Eusko Legebiltzarra. A number of smaller parties often play a pivotal role in determining which coalition governments can form. Further the parties are cleaved by multiple issues that cannot be boiled down to a simple left-right or conservative- socialist orientation. The electorate in the Basque country divides along at least two dimensions–a traditional Left-Right dimension, common in most European democra- cies, and a Nationalist dimension, ranging from complete incorporation into Spain to complete independence of the region from Spain (D´ıezMedrano 1995). The primary goal of this paper is to describe this setting and electoral choice in the region. Valence issues, such as the economy, and demographic and cultural features of the electorate also matter in vote choice, but the Left-Right and Nationalist dimensions cleanly cap- ture the main features of the party system in the Basque region and most of the voting behavior. An unusual aspect of the electoral and political system is the role of language. The Basque language, Euskerra, is often freighted with nationalist sentiments, especially

1The political conflict itself has been economically costly (Abadie and Gardeazabal 2003), so resolving the issue of autonomy and independence likely will have further benefits.

2 because both the Spanish and French governments outlawed the language at various times over the past 80 years. As we will show speaking Euskerra maps into electoral divisions among the region’s voters. However, the language question appears to have an independent effect on voting behavior from the effect of Nationalism, or the degree of autonomy. Language politics in the region can be separated from Nationalism, and should be viewed as identity politics. Identity politics, we argue, do not map neatly into the usual spatial model with valence issues, and this presents a opportunity to extend the spatial model to incorporate a non-position, but nonetheless spatial issue.2 The cultural and political divisions in the region would seem to make for a dan- gerously unstable political situation, especially given the history of civil unrest and violence during the 1930s and 1970s and 80s.3 In an electoral setting with multiple issues and multiple parties, political and social choice theory would typically predict highly unstable and shifting political coalitions.4 Empirically, M¨ullerand Strom (2000) and Diermeier et al. (2003) find that situations similar to those in the Basque Country typically have long-lasting negotiations to form a government, short-lived governments, and frequent changes in the party in power.5 Here lies the puzzle. Following the 10 elections in the BAC, government formation has taken relatively little time, with the average period between elections day and the investiture vote of 2 months. For most of its 35 years, the Eusko Legebiltzarra has been governed by one party, the Partido Nationalista Vasco in Spanish or Euzko Alderi Jeltzalea in Euskerra, commonly called the EAJ-PNV or just PNV. Cabinet members have only been modified once during the legislature (following the 1990 elections; see Table 2). Only from 2009 to 2012 was someone from a party other than the PNV chosen by the Legebiltzarra to be President of the Basque Autonomous Community.

2See Bourne (2010) describes the difficult political situation that nationalist parties are in, and Mart´ınez- Herrera (2002) study the effects of political decentralization on citizen identification with their region. 3See Joseba Zulaika’s excellent anthropology of the origins of the ETA uprising in the 1970s. 4Starting with Baron and Ferejohn (1989), many other authors have analyzed the difficult problem of inter-party bargaining from a non-cooperative game theory approach (see Ansolabehere et al. 2005, footnote 6 for additional references). 5See also Laver and Schofiel (1998) and Laver and Shepsle (1994).

3 The minority and coalition governments led by the PNV have been very stable and long-lived, and all but once have the governments lasted the full duration of the election cycle. A remarkable run of one party, minority government emerges out of a seemingly impossible political situation. The question is why. Why is there so much political stability in a region where the electoral and cultural circumstances ought to make for political chaos? The answer, we believe, lies precisely with the nature of identity voting.

2 The Party System in Basque Elections

The party system in Basque Parliamentary elections consists of a mix of regional parties and federal parties. The federal parties are the two primary parties in Spain that vie for control of the Cortes Generales as well as compete in regional elections. These are the Partido Popular (PP) and Partido Socialista Obrero Espanol (PSOE), which runs as the Socialist Party of the Basque Country (PSE) in the region’s elections. In the Cortes Generales, the PP and PSOE oppose each other, but in 2009 they managed to find common ground and formed a coalition government in the Basque parliament. Two other parties with presence in the national Spanish government also have a notable presence in the Basque regional elections. These are Izquierda Unida, IU or United Left and Union Progreso y Democracia (UPyD). IU runs in the Basque parliamentary elections as Izquierda Unida-Ezker Batua (IU-EB) and in recent years in league with the Green Party. UPyD is a liberal party that rejects Basque and Catalan nationalism, and was formed in 2007. It often aligns with the PP. The principal regional party is the PNV. Its main political platform is national autonomy, but on economic and social matters it promotes a very pragmatic platform. Various factions have split from the PNV and formed their own parties over time. The most significant fracture came in 1985 when then-Lehendakari Carlos Garaikoetxea Urriza clashed with PNV party leader Xabier Arzalluz over the direction of the party. As a result of the internal party struggle, Garaikoetxea founded Eusko Alkartasuna

4 (EA) in 1986, and he was replaced as president of the regional government by Jose Antonio Ardanza. The most controversial political parties in Basque regional politics are the leftist and nationalist parties organized under a variety of banners. Herri Batasuna (HB), or Unity of the People, formed in 1978. The party won 17,500 votes in the Basque and Navarre regions in the Spanish General election of 1979, enough to earn a seat in the Cortes Generales. But, under the Spanish constitution they were not allowed to claim that seat because of ties to ETA, especially ETA members running as candidates under the party label and ETA funding. In the 1980 regional elections in the Basque Country, HB received 17 percent of the vote – enough for 11 seats – but again was not permitted to sit in parliament. HB was renamed Batasuna (Unity) and ran with a second independence party Euskal Heritarrok, until Batasuna was banned in 2003.6 To fill the void left by HB/EH, the Communist Party of the Basque People reappeared in the 2005 election and won 12 percent of votes, and Aralar emerged attempted to capture the independence-oriented left voters, but it was never as successful as HB/EH. Bildu, a loose coalition of candidates and factions committed to independence, formed in the early 2000s and proved quite successful in local elections. In 2011 a number of leftist parties and political leaders who strongly support independence or greater autonomy – including Aralar, EA, Alternatiba, and some former Batasuna members – ran under a new party label Eukal Herria Bildu (EHB), or Basque Country Gather. In the 2011 Spanish general election EHB won a stunning 24% of the vote in the region, trailing only EAJ-PNV. The presence of candidates from Batasuna and Aralar in the EHB prompted a challenge to the legality of the new party in the run-up to the 2012 Basque Parliament election. Unlike three years earlier, the Spanish Courts did not , but the courts ruled that the party had not violated Spanish law and could run and hold seats in parliament. And in the 2012 election, EHB repeated its performance from a year earlier, winning 25% of votes and 21 seats in the Eusko Legeblitzarra. A secondary stream of the leftist nationalist parties were organized as the Euskadiko

6See Bourne (2010).

5 Ezkerra (EE), or Basque Country Left. In the first regional election in 1980 the socialist EE and communist Euskadiko Partidu Komunista (Basque Communist Party) parties won 14 percent of the vote. They merged in 1982 into a Marxist-oriented socialist, nationalist party, EE-IPS, but they never regained their electoral strength of the 1980 election. In 1991, the party split again. A majority of the party merged with the PSE. Adherents to the Euskal Ezkerra faction, however, joined with Eusko Alkartasuna. Table 1 shows the election results for every party since the first election to the Eusko Legebiltzarra in 1980. The table reveals both remarkable stability and tremendous instability in the Basque party system. The two principal parties, EAJ-PNV and PSE, have proven to be very stable forces. The PNV typically wins about 35 percent of the vote and wins on average 27 seats in the parliament. The PSE typically receives about 20 percent of the vote and 16 seats. The PP, although always present, has received less constant support. From 1980 to 1994, support for the PP was quite low in the region, in the single digits. The party surged from 1994 to 2001, but has since sunk back to about 12 percent of the vote. During the 1998 and 2001 elections, the PP supplanted the PSE as the number two party in the region. As a result, the Basque party system appears to be a 3+ system, with the plus added because of the chaos among the nationalist left parties. The nationalist left parties show tremendous instability. This surely owes to the re- peated bans placed on the ETA-related parties, Herri Batasuna, Batasuna, and Euskal Heritarrok. Other left parties, however, have not been able to find a solid footing in the region. The IU-EB/Green coalition has never cleared 10 percent of the vote, and the Communists come and go, depending on the presence of a strong independence party, such as HB. The splinter parties from the PNV – notably EA – were not able to maintain a presence and ultimately collapsed back into the PNV or reformed with other parties. The emergence of EHB in 2011, however, offers the possibility of a stable new party, the fourth for the region, and this leftist nationalist party looks like it could become the second largest party in the region, behind the PNV. Finally, it is worth noting that for much of the 35 year history of Basque parliamen-

6 tary elections there have been very few “wasted votes”– votes for parties that had no chance of clearing the threshold for receiving a seat. However, in the past two elections (2009 and 2012) more than 7 percent of all votes went to parties that won no seats. In 2012, 8 percent of the vote was distributed across many smaller parties, none of which won more than 5 percent in any province. In 2009, 9 percent of voters went to the polls and spoiled their ballots to protest the Spanish Court’s decision to ban Aralar and Batasuna. The formation of the government of the Basque Autonomous Community reflects the same odd mix of stability and instability as the parliamentary elections. Through- out the 35-year history of the parliament, the PNV has served as the governing party for all but 3 years (from 2009-2012), and until 2009, there had been only three differ- ent Lehendakari (Garaikoetxea, Ardanza, and Ibarretxe). From 1980 to 1990 the PNV held power as a minority government, a notably long tenure for a minority govern- ment. From 1990 to 1998, the PNV and PSE joined in coalition to form the regional government, along with several smaller parties. Over the next decade the PNV shed its alliance with the PSE, and formed coalition with EA, EH, and IU. Up until 2009, then, the government had always included the largest party in the region, the PNV, and that party chose the Lehendakari. But, in 2009, the negotiation to form a new PNV coalition broke down. First, the PSE insisted on selecting the next Lehendakari, a position unacceptable to the PNV. Then, the PNV failed to find common ground with the smaller parties of the left, primarily over differences in economic and development issues in the midst of the recession sinking the Spanish economy. In a stunning political maneuver, the PSE seized the opportunity to form a coalition with the PP – it’s adversary in the Cortes Generales. Uniting the two Spanish federal parties was their common opposition to the increasing autonomy of the BAC. The coalition of the PSE and PP was ill-fated from the start. It was brought about because the members of the EB decided to not vote in the selection of the government in 2009, and throughout its existence, the coalition hung on the vote of a single member of parliament. As the first year of the government

7 wore on, the differences between the PSE and PP on economic and social issues caused greater strains on their governing arrangement. Had the PSE and PP not formed a coalition in 2009, the PNV almost surely would have formed a minority government as it had from 1980 to 1990. Following the 2012 election, the PNV returned yet again to govern as a minority.

3 A Spatial Model

The electoral and party system in the Basque region can be understood analytically in terms of three types of issues confronting the Basque electorate. First, there are Positional or Spatial issues. These are issues over which there exists a policy choice, such as moving economic policy more to the Left or Right. The voters have distinct preferences along the dimensions that characterize each of the issues, and the parties adjust their platforms to compete for votes. Second, there are Valence issues. These are issues on which all voters are in agreement, such as economic prosperity and growth, and for which they hold elected officials accountable. Third, there are Identity issues. These are issues on which some groups of voters orient the same way, but other groups of voters may orient a different way. The voters classify the parties on the basis of an identity, such as race, language, or religion, but the parties cannot (at least in the short run) alter their identities. It is important to distinguish Identity from Nationalism. Nationalism is what Po- litical Scientists typically call a spatial issue. Nationalism involves a specific policy decision concerning the degree of autonomy of the region. The parties can alter their platforms concerning the amount of autonomy that they think the regional govern- ment ought to have. Identity is tied to long histories and symbolism that are not readily changed. Identity groups are extensively studied by social psychologists (see, e.g., Tajfel 2010). According to this literature, Ashforth and Mael (1989) suggest that “Identification induces individuals to engage in, and derive satisfaction from activities congruent with the identity, to view him or herself as an exemplar of the group, and to

8 reinforce factors conventionally associated with group formation”. Some people may never vote for HB, Aralar, or Bildu owing to family history or events during the ETA uprising; others will only vote for those parties. Still others may only vote for nation- alist parties because of their identity as Basques, rather than Spaniards. The parties cannot change their identities or the identities of the people. Unlike Valence issues, Identity issues do not affect all people the same way. Unlike Spatial issues, the parties cannot easily adjust their image on Identity issues. As is well known, pure strategy Nash equilibria exist only under special circum- stances for multi-party elections in a multi-dimensional setting. When those condi- tions do not adhere, chaos results. However, the valence and identity issues broaden the conditions for finding equilibria. The purpose of this paper is not to characterize the equilibria in the electoral setting, but to use this framework to help us analyze the politics in the region. Our intuitions about the characteristics of likely equilibria derive from valence politics models in multiple dimensions with complete information (Ansolabehere and Snyder 2000; Aragon`esand Palfrey 2002; Schofield 2003; Schofield and Sened 2005). Schofield (2003) introduces the term “activist valence” by which the activist members of the party contribute to increase the valence of the party in exchange for moving the platform of the party closer to their more extreme position. The behavior of the electorate under multiple dimensions with “activist valence” is quite similar to that of the electorate with Identities. Mathematically, we represent electoral choice in this setting as follows.

3.1 Preferences of voters over positional issues

There is a society with a continuum of voters which shall select by popular election a representative to serve in the legislature. There are two main positional issues in the society, the ideological issue (issue 1) that is measured by the left-right scale and the nationalist issue (issue 2 ) that measures the support for policies ranging from complete independence to complete incorporation into Spain. These two issues define

9 the policy space.

There are N political parties. Each party j takes a position xj = (xj1, xj2) where

xjr is the political position of party j on issue r ∈ {1, 2}.

Each voter i has an ideal policy: πi = (πi1, πi2) where πir is the ideal position of voter i on issue r. Spatial preferences are usually represented by the Euclidean distance according to the following utility function

2 2 ui(j) = −α1[xj1 − πi1] − α2[xj2 − πi2] − α12[xj1 − πi1][xj2 − πi2] (1)

where, for each issue r, αr > 0 indicates the weight that voters assign to issue r and

α12 ∈ R indicates the weight that voters assign to the interaction between issue 1 and issue 2 (this last term can be either positive or negative).7 Indifference curves are

ellipses centered in the ideal point of the voter. In the case that α12 = 0 and α1 = α2,

indifferent curves are perfect circles and when α12 = 0 and α1 > α2 , the saliency of issue 1 is higher than that of issue 2 and indifference curves become narrower. Figure 1 illustrates two examples of voters’ indifference curves.

Let j and k be two distinct political parties. Voter i is indifferent between the

two parties when ui(j) − ui(k) = 0. Solving for πi2 in this equation, we obtain the expression that describes the locations of the indifferent voters:

πi2 = a − bπi1 (2) 2 2
2 2
α1 x − x + α2 x − x + α12(xj1xj2 − xk1xk2) a = j1 k1 j2 k2 (3) 2α2 (xj2 − xk2) + α12 (xj1 − xk1) 2α (x − x ) + α (x − x ) b = 1 j1 k1 12 j2 k2 (4) 2α2 (xj2 − xk2)

This expression is a line which always contains the midpoint between the positions of

party j and party k. Figure 2 provides two examples. In the first one, α12 = 0 and

α1, α2 > 0 and the line always has a negative slope (in particular, when α1 = α2 this

7Other distance functions can replace the euclidean distance when representing the preferences of voters according to the spatial model.

10 is orthogonal to the vector xj−xk). In the second example, we illustrate that when

α12 6= 0, the line can show a positive slope.

3.2 Positional, valence, and identity issues

Political parties are also endowed with valence issues and identity issues. Valence issues, such as economic growth, are attributes or evaluations of the parties equally valued by all voters. Identity issues are attributes of the parties that are valued dif- ferently by different groups within the electorate. Language and race may function as identity issues in many societies. Consider the simplest case in which, according to their identity, the electorate can be partitioned into two groups (e.g., dividing into those who want to preserve Euskerra and those who do not care about this). Let P = {E,S} denote a partition of the electorate. When positional, valence, and identity issues are incorporated, each political party is

11 characterized by a profile (xj, vj, dEj, dSj) where dEj, dSj ∈ R measure how voters belonging to the identity groups E and S respectively feel about party j. The utility representation of preferences of voters when i ∈ E is

2 2 ui(j) = vj + dEj − α1[xj1 − πi1] − α2[xj2 − πi2] − α12[xj1 − πi1][xj2 − πi2] (5) and when i ∈ S

2 2 ui(j) = vj + dSj − α1[xj1 − πi1] − α2[xj2 − πi2] − α12[xj1 − πi1][xj2 − πi2] (6)

The set of voters that are indifferent between Party j and Party k depend on the voters’ identity group. Solving for ui(j) − ui(k) = 0 we deduce:

πi2 = vk − vj + dEk − dEj + a − bπi1 when i ∈ E (7) πi2 = vk − vj + dSk − dSj + a − bπi1 when i ∈ S

12 where a and b are as defined in (3)and (4). Identity issues may reinforce or reduce the effect of valence in voter’s preferences within each identity group. Figure 3 illustrates

this point. Consider that 0 < dEj < dEk and 0 < dNk < dNj, i.e., voters in group E feel closer in identity to Party k whereas voters in group S feel closer in identify to Party j. Observe that, in contrast to Figure 2, voters with equal or close ideal points may prefer different political parties. For this to be the case, these voters must belong to different identity groups.

Figure 4 describes the case of three political parties when there are identity issues. In the upper left graph, we represent the policy space divided into three areas each of which corresponding to the location of the ideal policies of voters with common top-ranked political party.8 The two graphs on the right analyze how the three areas

8 For simplicity we consider that α1 = α2 α12 = 0 and vj = vk = vl. In this case, the ideal point of the voter that is indifferent among the three parties corresponds to the circumcenter of the triangle with vertices in the parties’ policy positions.

13 change for each identity group when 0 ≤ dEj < dEl < dEk and 0 ≤ dSk < dSl < dSj. We first analyze the identity group E. The area that contains the location of those voters that prefer Party k over the two other parties expands and the area of those that prefer Party j over the two other parties shrinks. There is an ambiguous effect regarding the area of those that prefer Party l over Party k and Party j. For the identity group S, the area that contains the location of those voters that prefer Party j over the other parties expands, and the area of those that prefer Party k over the two other parties shrinks. When the two groups are analyzed together, we obtain an inverted Y-shaped area that contains the location of the ideal points of those voters that, depending on their identity group, prefer different political parties.

14 4 Mapping the Basque Electorate

Basque parliamentary elections conform remarkably well to a spatial model character- ized by positional issues and identity and valence characteristics. Using public opinion surveys of the BAC region, we can map the preferences of people and the positions of the parties, and we can gauge the extent to which the various types of issues shape voting behavior. The Centro de Investigaciones Sociologicas (CIS), an independent entity of the Ministry of the Presidency of Spain, conducts surveys during national and regional elections throughout the country.9 CIS began conducting national sociological surveys in 1963, and their surveys cover all regional and national elections in the Basque Country from 1980 on. Households are selected at random and the interviews are in person. The surveys are conducted before and after the election. The typical sample size is 1,400. We pool the pre- and post-election surveys when both are available, yielding samples of 2,800. The increased sample size helps with the estimation of vote preferences, especially for smaller parties. We study the CIS surveys from 1994 to 2012, as these surveys contain appropriate measures for examining the importance of ideology and nationalism in Basque parlia- mentary elections and the Basque party system. The surveys ask people whether they voted and how; various demographics, such as age and gender; sociological character- istics, such as languages spoken and cultural identities; and political attitudes. These indicators allow us to gauge the nature and importance of spatial voting, identity vot- ing, and valence voting. Our analysis will focus on five key variables from the survey: Vote Preference or Choice, Nationalist Orientation, Left-Right Orientation, Basque Identity or Language, and Assessment of the Economy. Not every survey contains all of the indicators of interest. Vote Choice or Preference is the outcome of interest. The surveys branch the voting questions, asking people first whether they voted (or planned to vote). Of voters (or

9www.cis.es.

15 likely voters) the survey asks for which party or coalition of parties the individual voted. Left-Right Orientation measures the ideological position of the person. “Normally when talking about politics the expressions left and right are used. On this cared there are a series of boxes that go from left to right. In which box would you place yourself? The box 01 is labeled “Izquierda” for left and 10 is labeled “Derecha” for right. The second dimension of interest is Nationalism. The survey asks ”In relation to the nationalist sentiment, could you tell me please where you would place yourself on a scale from 1 to 10, in which 1 means the least Basque nationalism and 10 the most Basque nationalism?” We use these questions to map out the ideological orientation of individuals. In addition, four of the surveys (1998, 2005, 2009, and 2012) ask respondents to place the parties on the Nationalist and LeftRight scales. We use responses to these questions to measure where the parties are in the two-dimensional space and the stability of their policy positions. Separate from nationalist preferences, the CIS surveys asks various questions that gauge identity. Most of the surveys ask whether the individual speaks Euskerra fluently. In addition, the surveys ask whether people identify themselves as only Basque, more Basque than Spanish, equally Basque and Spanish, more Spanish than Basque, or only Spanish. We use these questions to gauge the identity of individuals and how Identity issues explain vote preferences. We also include a indicator of the size of locality that the person lives in and the Province, as studies of aggregate voting patterns conclude that town population correlates strongly with nationalist party vote. Finally, survey respondents evaluate the state of the economy in the Basque country. “What is your view of the economic situation in the Pays Basque today? Very Good, Good, Average (Regular), Bad, Very Bad” captures the most common form of valence issue, economic voting.

16 4.1 The Electorate’s Preferences

The CIS survey provide a clear picture of of the Left-Right and Nationalist orientations of the Basque electorate and party system. Turning first to the electorate, we can map the positional issue preferences of Basques along each dimension separately and in a two-dimensional space. Tables 3 through 6 characterize the spatial preferences of voters. The ideological orientation of the Basque electorate is highly Centrist, with a slight left of center cant. Pooling the surveys from 1994 to 2012, we find that the modal ideological identification is 5 – 30 percent of adults place themselves exactly in the center of the scale. Another 18 percent chose 4 and 21 percent chose 3. Over three- quarters of respondents placed themselves in the interval from 3 to 6 on the scale. Fifteen percent chose the far Left (either 1 or 2), and only 8 percent chose a position to the right of Center (7 to 10). The median voter along the Left-Right spectrum places herself at 4. Pooling all years, the median is 4 and the average score on the 10 point Left-Right scale is 4.2. The high centrist concentration of the electorate is reflected in the relatively small standard deviation of just 1.7. Moreover, the distribution of preferences along the Left-Right dimension have been very stable. The average, median, and standard deviation have not changed in any meaningful way over the past 20 years. (See Table 3.) On questions of Nationalism, the Basque electorate also appears fairly centrist, with a tilt in favor Nationalism and greater regional autonomy. Again pooling the sur- veys from 1994 to 2012, we find that the modal response to the Nationalism question is exactly in the center, with 20 percent choosing 5 on a scale from 1 (Minimal Na- tionalism) to 10 (Maximal Nationalism). The distribution of preferences, in contrast with Left-Right ideology, is not concentrated around the center, but is quite polarized. Forty percent of respondents support greater Nationalism and autonomy (7 to 10 on the scale), while a quarter (25 percent) support minimal Nationalism (1, 2, or 3 on the scale).

17 The median voter along the Nationalist scale places herself exactly in the center. The median placement is 5, and the mean is 5.6 on the Nationalism scale. Unlike the Left-Right scale, the Nationalism distribution is more widely dispersed around the mean, as reflected in the standard deviation of 2.8. Nationalist attitudes have also exhibited some trending over time. In 1994, the average Nationalist score was 6.3, but by 2012 it had fallen to 5.0. In addition, the spread of the distribution has increased. In other words, the center of Basque electorate has shift from somewhat Nationalist to Moderate on the question of Nationalism, but the degree of polarization on this issue has also increased. The standard deviation of the Nationalist Scale was 2.4 in 1994, and it rose to 3.1 by 2012 – a thirty percent increase in the dispersion of the electorate away from the center on the question of Nationalism. (See Table 4.) In 1994, those who supported greater national autonomy outnumbered those who supported minimal nationalism by 3 to 1. Today, these groups are about equal is size, and each account for slightly more than a third of the electorate. Interestingly, the Left-Right and Nationalist dimensions of Basque politics appear to be unrelated to each other. There are historical reasons why one might expect some association between Ideology and Nationalism. During the Spanish Civil War, for ex- ample, Russia supplied arms and tactical support for the guerrilla fighters loyal to the monarchy, while the U.K. and United States stayed on the sidelines. This had a radi- calizing effect on those fighting to defend the nascent Basque Republic. Similarly, the organizations associated with ETA in the 1970s and 1980s aligned very strongly with Marxism and revolutionary ideologies. After Franco’s death, his supporters aligned most strongly with the PP in the Cortes Generales and in the regional elections. His- tory, it would seem, laid the foundations for alignment of nationalists with the left and of pro-Spain factions with the right. This feature of Basque history certainly is reflected in parties such as EH and HB, but it neglects the more conservative strain of nationalism. In particular, the Basques were loyal to the king in the Civil War, because the royalty had long granted them rights of independent governance, and the Basque region is among the most religiously Catholic areas of all of Europe.

18 In the Basque region, people’s views exhibit only a slight negative correlation be- tween Nationalism and Ideology of -.14. That correlation has varied somewhat from year to year, but there is no clear trend of either weakening or strengthening ties be- tween Nationalism and Ideology in the Basque electorate. There is a slight, noticeable correlation, but it is not the strong association one might expect from the annals of Basque and Spanish history. We capture the relationship between pro- and anti-Nationalist sentiment and be- tween Left and Right ideology in Table 5. This table distills the 10-point scales down to a simpler representation of Ideology as Left-Center-Right and of Nationalism as Minimal-Moderate-Maximal. We collapse the values 1, 2, and 3 on each scale to indi- cate those on the Left and those on the Minimal Nationalism end of each spectrum. We collapse the values 4, 5, and 6 to indicate Centrists and Moderates. And, we collapse the values 7 through 10 to indicate those on the Right and Nationalists. Reading across the rows of the table one can see that most people are Centrists. Reading down the columns one can see much more dispersion of people’s preferences about Nationalism. The degree of centrism, though, is quite clear. Almost a quarter of all people in the sur- veys identify as Centrist-Moderates, and another 20 percent at Centrist-Nationalists. The map of the Left-Right and Nationalist orientations of the Basque electorate is quite informative about what one might expect of the electoral alignment of the parties. In a proportional representation system, such as for the election of the Eusko Legebiltzarra, it would make sense for at least one party to occupy the centrist position, as that is where there is the greatest density of voters. The Nationalist Center and Nationalist Left would also seem to be strong bases of electoral support.

4.2 Perceptions of the Party System

The party system, as our discussion in section 2 suggested, is quite stable, and the stability is reflected not only in the vote shares of the parties but in the policies and platforms that the parties present to the electorate. The CIS surveys allow us to gauge

19 the positions of the parties along the same Left-Right and Nationalist scales as the electorate. The surveys ask respondents to place the parties on each of the two scales, as well as themselves. The average party score (from the voters’ perspective) on the Left-Right and Nationalist dimensions are shown in Table 7. The placements of the parties provide a mapping of the party system that is con- sistent with descriptive accounts of the Basque party system. The four major parties or coalitions have staked out quite distinctive electoral bases. The EAJ-PNV, the largest party in the region, is strongly Nationalist, with a Nationalism score of 8.2, and somewhat Right of Center, with a Left-Right score of 6.4. The PSE appears clos- est to the center. It is slightly Left of Center, with a Left-Right score of 4.5, and somewhat Anti-Nationalist, with a Nationalist score of 3.4. The PP is Right and Anti- Nationalist, with a Left-Right score of 8.6 and a Nationalist score of 1.9. The string of parties with labels HB, EH, and Bildu occupy similar ideological space, being highly Nationalist (score of 8.7) and farthest to the Left (Left-Right score of 2.1). In sum, the 4 main parties or coalitions occupy the Nationalist-Center, the Nationalist-Left, the AntiNationalist-Center, and the AntiNationalist-Right Smaller parties – notably IU, EA, and UPyD – also occupy important positions within the system. IU is quite moderate on the Nationalist dimension and farther to the Left than the Socialist Party. IU-EB increasingly aligns with the Greens. EA, which split from the PNV in the 1980s, is Centrist on the Left-Right dimension and strongly Nationalist. One might think that EA would have a superior electoral position to the PNV, because EA is closer to the center ideologically. Finally, UPyD, and a similar faction UA, occupy a Conservative-AntiNationalist position. UPyD is slightly more moderate than the PP and is a natural coalition partner for the PP. Interestingly, even though EA and UPyD are closer to the center than their partner parties (PNV and PP) they have not established strong electoral support. The situation of EA is particularly puzzling from this perspective. They appear to occupy a relatively strong electoral position, but, as we shall see, EA never established itself as a strong electoral faction to rival the PNV.

20 A few comments about the overall party alignment are in order. First, the parties stake out very distinct policy positions. The PNV, PSE, and PP occupy very different policy areas within the two-dimensional space. The PP and UPyD occupy the Nationalist Right. The PNV and EA have staked out the Nationalist Center and Center Right. The IU and the PSE appeal to the Left and Moderate or anti-Nationalist voters. Even within each of these parts of the electorate the parties manage to distinguish themselves. Second, the party alignment in terms of positional issues is extremely stable. There is almost no variation from year to year in the positions of the parties. In a mutiparty, multi-dimensional setting one might expect more instability, but even the smaller par- ties exhibit very little movement over time. What is more, the party system remains stable even after some of the parties are banned or fold. Herria Batasuna was banned in 2003, but no of the incumbent parties shifted to capture the substantial Left-Nationalist vote. When HB is reformed into Bildu it occupies exactly the same ideological location as HB, at least in the voters’ minds. Third, there is no centrist party. The PNV is often described as a moderate or Center-Right party, but the electorate sees it as on the extreme in terms of Nation- alism and noticeably to the Right of the typical voter ideologically. EA is equally nationalistic, but centrist ideologically. The PSE is close to the center of the Left- Right spectrum, but favors minimal nationalism. IU is centrist on Nationalism, but far to the left ideologically. The electorate looks quite different, as was shown in Table 6. The median voter in the region is Centrist or slightly Left of center and Moderate or slightly pro-Nationalist. There is no party in the system that represents this pair of positions. The closest to the center might be EA or the PSE, and these parties, as we see in the next section have failed to capture the upper hand in Basque parliamentary elections.

21 5 Spatial Voting

How well do positional, identity, and valence issues account for patterns of voting in Basque parliamentary elections? It is useful to keep in mind the positions of the parties and the density of voters in different segments of the two-dimensional ideological space. We expect to see people to vote for parties that have the same orientation on the Left- Right Scale and those with the same Nationalist orientation as them. We also expect that assessments of the economy and Basque and Spanish identities will pull people in various directions. Here we offer an assessment of the strength of the appeal of various types of issues and a comparison of the appeal of Left-Right and Nationalist ideals. The pure spatial model (with only positional issues) carries a fairly stark set of predictions about which people vote for which parties. A person only votes for the party that is closest ideologically. Generally speaking, closeness depends on the orientation and steepness of the contours of voters’ preference functions. For simplicity assume that voters weigh both issues equally and hence have spherical utility curves. Then, simple distance in two-dimensions determines proximity. That model provides a remarkably powerful description of how people vote in Basque parliamentary elections. A voter Nationalist-Leftist voter will choose HB or Bildu and not vote for the PNV or PSE. An Anti-Nationalist-Rightist voter will choose PP (or perhaps UPyD) rather than PNV or PSE. Identity politics complicate this picture somewhat. As discussed earlier, it is possi- ble for identity politics to push some voters away from parties that they would otherwise support. Consider, for example, an individual who identifies as Right of center and who supports greater autonomy for regional governments as a matter of principal, but the individual who also identifies as Spanish, rather than Basque. That individual might vote for the PP because that party is Spanish and the nearest alternative, the PNV, identifies as Basque. If identity is uncorrelated or negatively correlated with ideological or nationalist orientation, then identity can produce a significant vote that blurs the lines establishing positional voting. If identity is positively correlated with

22 positional issues, then identity can strengthen voting on the positional issues. For ex- ample, Basque identity is positively related to preferences on regional autonomy; people who speak Basque fluently favor separation. This positive relationship is expected to increase the appeal of regional autonomy among those who identify as Basque and increase the appeal of stronger central government control over the region among those who identify as Spanish.

5.1 Votes

We wish to explain which party a given individual chooses to vote for. The survey measures voters’ preferences in the pre-election polls and choices in the post election polls. In any given year the pre- and post-election surveys are very similar, so we feel justified in pooling the two surveys. Table 8 presents the percent of people in the CIS survey who said they intend to or actually voted for each of the parties. One concern with any election survey is the degree to which the responses accurately reflect actual behavior. Compare Table 8 with Table 1. In every year, the share of the reported vote for the PSE, HB/EH/Bildu, and IU parties in the survey is quite close to the actual vote for each of these parties. The estimated support for the these parties in the survey is never more than a couple of points off of the parties’ actual performance. The most glaring discrepancies in the poll arise with the PNV and the PP. On average the CIS surveys overestimate the vote for the PNV by 8 percentage points, and understate the support for the PP by 9 percentage points. The poll results for the PNV are reasonably close to the party’s actual performance in 2009 and 2012, but there are large discrepancies between the actual and estimated vote in 1994 and 2001. The correlation between the survey errors for PP and for PNV is -.49. We are unsure why these errors arise; we do not think these errors affect our inferences appreciably, but they may merit further investigation. It is worth emphasizing the impressive showing of the PNV, PSE, and PP. These are the three main parties in Basque elections. They are somewhat more extreme than

23 the smaller parties with whom they share political common ground – EA in the case of PNV, IU in the case of PSE, and UPyD in the case of PP. However, the PNV, PSE, and PP do much better than the smaller parties. Most notably, in years that it ran, EA, which is ideologically closest to the center, never managed more than 15% of the vote, and its support shrank over time.

5.2 Positional Issues and Vote Choice

Issue voting in Basque elections can be thought of as a calculation made by each voter as to which party is closest to their ideal points. We have a crude map of party positions and voters’ ideal points based on placements on the Nationalist and Left- Right dimensions. Imagine overlaying Table 7 (the positions of the parties) on Table 5 (the distribution of positions of the voters). Table 7 tells us the ideological location of each of the parties in the two-dimensional space. Table 5 shows us where the density of voters are in that space. In the absence of valence and identity issues, a party will win all of the votes in a given cell of Table 5 if there are no other competitors in that cell. If two or more parties occupy a cell they will split the votes. If no party occupies a cell of the table, the parties closest to that cell will split the votes. For example, if a voter is a 4 on the Left-Right scale and a 3 on the Nationalist scale, he or she will vote for the PSE, because that party is closest, and not for the PP or PNV or any of the other parties. And, all voters at the location [4,3] will vote for the PSE. If 5 percent of all voters have those policy preferences, the PSE wins that 5 percent of the votes. If positional issues strongly determines vote choice, then voters ought to sort themselves fairly cleanly along these lines. Table 9 shows the vote for each party in each sector of the two-dimensional space spanned by Ideology and Nationalism. Starting with the classification of voters in Table 6, we calculate the percent of all survey respondents in each cell of the table who chose a particular party. For example, the first panel inside Table 9, presents the vote for the PNV for each subgroup created by crossing Ideology and Nationalism. The first

24 cell indicates that 5.2% of people who are Leftists and Anti-Nationalist voted for the PNV. The remaining 95% of respondents who identify themselves as Anti-Nationalist Leftists voted for other parties. This table excludes non-voters. The table shows a high degree of spatial voting along Ideological and Nationalist lines: People vote for the party closest to them. The base of the PNV are Nationalists who are Center or Right on the Ideological spectrum. Consider people who identify as Right of Center and Nationalists, the bottom right cell in each table. Over 90 % of those people, across all elections, voted for the PNV. The rest of the vote of these people was scattered: EH/HB/Buildu captured 4% and the PP, 2%. Now consider people who are ideological Centrists and Nationalists. The PNV won 86 % of the votes of these people. Center and Right Nationalists, then, are the base of the PNV, and the PNV wins almost all of the voters who hold these attitudes. The PNV also won 31% of votes of people who consider themselves Right of Center ideologically and Moderate on the Nationalism question; the party won 31% of votes of people who are Left of Center and strongly Nationalist, and the party won 51% of voters who consider themselves Centrists and Moderates. This pattern fits perfectly with what one would expect from positional voting in a spatial model. The party is somewhat Right of Center ideologically and strongly Nationalist. It has no real competitor parties in that space, and it wins nearly all of the Right and Center votes among those who are Nationalists. It wins half of the Centrist-Moderate vote, and there are no parties in that space. And, it makes a substantial second choice among those who are Left-Nationalists and Right-Moderates. Turning to the Anti-Nationalist Right, we find the base for the PP and UPyD. The PP won 68 percent of all votes cast by people who describe themselves as Right of Center and opposed to Nationalism. The UPyD received votes of 20 percent of these people. The PP also won 45 percent of votes of people, and the UPyD received the votes of 9% of these people. Support for these parties drops precipitously as one moves to the left or in the Nationalist direction. The Moderate and AntiNationalist Left is reflected in the support for the PSE and

25 IU-EB/Green parties. The PSE receives the support of 65% of Leftist-AntiNationalists; 55% of the vote of Moderate-AntiNationalists; a third of the votes of Leftist-Moderates; and 27% of Centrist-Moderates. In other words, as one moves away from the Leftist- AntiNationalist pole, support for the PSE drops quickly. The IU-EB/Green coalition comes in a distant second among Leftist-AntiNationalists, with 19% of their votes. And, this coalition competes with the PSE for the support of the Leftists who are more moderate on Nationalism, winning 30% of these votes. Beyond that, the IU-EB receives very little support. Finally, consider the Left and Nationalist parties EH, HB, and Bildu. These parties ran in 1994, 1998, 2001, and 2012. In those years, the parties won 61% of the votes case by people who considered themselves Left and Nationalist. They won 15% of the votes of Left leaning people who were otherwise Moderate on the Nationalism question and 9% of votes of Leftists who were AntiNationalist. Their appeal to the Nationalist - Centrists, however, was weak, and they won only 7% of these people’s votes. We did not show the vote for EA, as that party received only a small share of the votes among survey respondents. The party is Centrist ideologically and strongly Nationalist. It received only 12% of the Center Nationalist vote, which is anemic compared with the PNV’s 78% among these voters, and among the Leftist Nationalists EA won only 7% compared with 19% for the PNV among these voters. Why EA’s vote was so low is odd, given their ideological moderation. One possibility is that the party was exceedingly close to the 5% threshold, below which a party does not receive a seat. Supporters of EA might have strategically chosen PNV in order to avoid wasting votes if the party did not reach the threshold for winning a seat. Whatever the explanation for EA’s weak showing, the degree of sorting of voters along ideological and nationalist lines suggests that positional issues offer a powerful explanation of Basques’ electoral choices.

26 5.3 Statistical Model of Vote Choice

The model presented in the third section of this paper provides a framework for un- derstanding the link between the mapping of the party system and voters’ decisions. Recall that there are four features of the model that describe which voters prefer which parties. First, there are the parties’ platforms – the location of the parties in the space defined by Ideology and Nationalism. Second, there are the preferred policies of voters – the distribution of voters’ ideal points in that two-dimensional space. Third, there is the intensity of voters’ preferences – the weight that voters place on one dimension over another. Fourth, identity and valence issues pull voters away from their ideologi- cal and nationalist orientations. Here we measure the weight of Ideology, Nationalism, and Identity and Valence Issues in voters’ decisions. The positional issues in that model imply that the choice between any two parties can be characterized as the distance of an individual voter from one party compared with another party. We can further define the set of voters who are indifferent between any two parties: those on one side of that line (say on the side of party A) prefer party A, and those on the other side of that line prefer party B. With three or more parties we can define the lines that separate each pair of parties to construct the set of voters who prefer a given party, say A, over its competitors. The probability that a voter chooses party A over party B is P (U(A) > U(B)). Suppose that the dimensions are Y and X and that the voter i gives weight of α to issue X and β to issue Y . Let the valence issue be w and the net identity term be di. Then,

2 2 2 2 P r(V = A|AorB) = P r(α(xA−xB−2(xA−xB)xi)+β(yA−yB−2(yA−yB)yi)+w+di > 0)

The terms xA, yA, xB, and yB are the positions of parties A and B on issues X and

Y . This may be rearranged into a probability function whose argument is linear in xi and yi, w, and di. That is,

2 2 2 2 P r(V = A|AorB) = P r([α(xA−xB)+β(yA−yB)]−2α(xA−xB)xi−2β(yA−yB)yi+w+di > 0)

27 Hence, for any paired comparison between parties A and B, the choice function can be modeled as a linear function of the voter’s position on X, the voter’s position on Y , the valence term w, and the identity term di. We will use these relationships to estimate the approximate salience of each of the dimensions. A logit or probit model may be used to estimate the probability that a respondent choose party A over party B as a function of xi (Ideology), yi

(Nationalism), w (economic circumstances), and di (identity). It is tempting (and common) to interpret the coefficients on xi and yi as the salience or weight of the issue in the voter’s thinking or utility function, but the equation above reveals that interpretation is incorrect. The coefficient on the voter’s position on a given issue dimension is a function of the salience or weight, but it also depends on the distance between the parties. Specifically, the coefficient on xi and yi is 2 times the distance between the parties on that issue (i.e., (xA − xB) or (yA − yB)) times the salience of the issue (i.e., α or β). We have measures of the distance between each pair of parties on the various dimensions from the analysis in the prior section. The PNV and the EH/HB/Bildu are quite similar on Nationalism and quite different on Left-Right Ideology. If the coefficient on Nationalism and Ideology are approximately the same for the comparison of the PNV versus the EH/HB/Bildu, then the salience of Nationalism must be quite high relative to the salience of ideology. We can recover the salience of the issue by dividing the logit or probit coefficient by 2(xA − xB) or 2(yA − yB), depending on the issue. We estimated this model using a multinomial logit model for each year, as the parties running for election vary from year to year. For each party we estimate the probability of voting for that party or for the PNV (e.g., PSE v. PNV). That is, party B in our analysis is always the PNV, and party A is one of the other parties. Since the PNV is the governing party in all elections except for 2012 (when the PSE held the position of Lehendakari), we expect any economic voting to be a retrospective assessment of the PNV from 1994 to 2009 and a retrospective assessment of the PSE in 2012. Our measure of National Identity is whether the individual speak Euskerra.

28 Finally, size of community (Population) is included in the analysis as that is argued to be an important predictor of voting by past research. Table 11 presents the estimated coefficients and standard errors from this analysis. We perform a separate analysis for each year, so each panel should be read across the table. There is clear evidence of all three types of issue voting – valence, identity, and positional issues. Valence issues appear to have the smallest effect in magnitude. There are strong effects of economic assessments across the board in 2001 and 2005, but no significant effects in 2012. We will set aside such retrospective judgments as they appear to be of more marginal importance. Similarly, community size does not have regular and predictable effects. Population has some significant effects, especially contrasting the PP and PNV, but the effects are highly unstable from year to year and most comparisons are insignificant. The important factors that emerge in our statistical analysis are Identity, Ideology, and Nationalism. Population, the economy, and other demographics are of secondary importance. Valence issues, measured as the respondent’s subjective evaluation of the economy in the Basque country, have minimal effect on voting in this region. The coefficients on economic assessments are significantly different from 0 in only 5 of the 15 comparisons for which we could estimate an effect (there was no economy variable in the 1998 survey), and the economy played no significant role in distinguishing the PNV and the PSE. It mattered most in voting for the Leftist-Nationalist parties PCTV and EH. Population was also rarely an important indicator of preferences. The important explanatory variables were Euskerra speakers and individuals’ preferences along the Nationalism and Left-Right scales. Consider, first, the role of identity, measured as whether the individual speaks Euskerra. We consider this to be an Identity indicator because it is the most obvious and salient indication of identity in this region, and it appears to have a separate effect from Nationalism. Inclusion of an indicator as to whether the person speaks Euskerra did not change the coefficient on Nationalism appreciably. Nationalism is a policy choice; speaking Euskerra is a characteristic. The parties themselves are tied

29 to identity politics. Euskerra speakers, holding constant ideology, nationalist views, population, and economic performance, are much more likely to vote for the PNV over the PSE, PP, or IU. The coefficient on speaking Euskerra appears to be strong and fairly steady over time. Euskerra speakers strongly prefer Bildu, EH, and the PCTV to the PNV. Consider difference between support for the PNV or the PSE. Although there is some shifting in the coefficient from year to year, the average coefficient is .98. That implies that the difference between an Euskerra speaker and a Spanish speaker is approximately a 10 percentage point difference in the probability that someone votes for the PNV over the PSE. The average coefficient for the PP versus the PNV is even larger, about 1.20. That translates into roughly a 12 percentage point difference in the probability that an Euskerra speaker votes for the PNV over the PP. (Note: the difference arises because of the lopsided distribution of votes for PNV over PP in the sample.) The largest effect of identity of the probability of support, by far, is for the EH/HB/Bildu parties. The probability that an Euskerra speaker voters for EH/HB/Bildu is 21 points higher among Euskerra speakers than among Spanish speakers. Language, then, operates as a valence, drawing Euskerra speakers toward the Nationalist parties relative to Spanish parties. Positional issues – Ideology and Nationalism – have even larger effects on voting. The overall effect of a variable can be measured as the change in the probability of choosing one party over another for a one standard deviation difference in the inde- pendent variable, holding other variables at their mean value. That calculation reveals that the overall effect of Nationalism on vote is approximately the same, or perhaps greater than, the overall effect of Ideology. This is because of the greater degree of polarization along the Nationalist dimension than along the Ideological dimension. A one standard deviation change on the ideology scale is only a change of 1.6 units, but a one standard deviation difference across people in the Nationalism scale is 2.8 units along the scale. Hence, the total effect of differences in Nationalist attitudes in the electorate as a whole is approximate 3 times larger than the estimated coefficients would indicate, and about twice as large as a one standard deviation difference in the

30 Ideology variable. Standardizing the variables and reestimatng the models reveals that the standardized coefficient on Nationalism is typically. For 2012, for example, the standardized logit coefficient on Nationalism is -1.70, compared to a raw coefficient of -.57, and the coefficient on the standardized ideology measure is -1.34, compared with -.80. A one standard deviation difference on the Nationalist scale accounts for approximately a 50 percent change in the probability of voting for the PNV, where a one standard deviation difference on the Left-Right scale amounts to a 30 to 40 percent difference in the probability of voting for the PNV. The overall effect of an ideological variable, such as Left-Right or Nationalism, reflects the degree to which the electorate divides between the parties along those lines when voting. The overall effect of a variable does not measure directly the salience or weight of the issue. In fact, many of the coefficients are negative, and it is not clear what negative salience would even mean. However, we can derive the underlying salience parameter using the coefficients estimated in Table 11 and the figures in Table 7. It should be the case the the salience parameter implied by the estimated logit parameters are about the same from party to party for a given issue, but may be different for Left-Right and for Nationalism. We calculate an overall estimate of the salience parameters for Nationalism (α) and for Left-Right (β). First, compute the average position of the parties over time as the average of the rows in Table 7, and then calculate the deviation of a given party from the PNV. Second, compute the average value of the logit coefficients for a given party on a given issue across time. The average coefficient divided by negative two times the deviation of the party from the PNV estimates the salience. The estimated salience of Left-Right ideology is .18 in the comparison of the PSE and the PNV, .23 in the comparison of the PP and the PNV, .14 in the comparison of the IU and the PNV and .16 in the comparison of the PNV and the EH/HB/PCTV/Bildu. Notice how similar the salience term is from party to party and that the salience term is always positive; that is consistent with our theoretical model. Now consider nationalism. The estimated salience of Left-Right ideology is .08 in the comparison of the PSE and the

31 PNV, .08 in the comparison of the PP and the PNV, .13 in the comparison of the IU and the PNV and .08 in the comparison of the PNV and the EH/HB/PCTV/Bildu. As with Left-Right ideology, the salience coefficients are positive and very stable from party to party. A few comments about the salience estimates are in order. First, in nearly every model, the salience of Left-Right Ideology is substantially higher than Nationalism, about twice as large. This implies that voters in the Basque region care much about Left-Right ideology than they do about Nationalism, and that has been true for the past two decades. Second, the average salience of Nationalism is .09 and the average salience of Left- Right ideology is .18. Third, the salience terms are very similar from party to party, even from year to year and party to party. This suggests that the model of issue salience as a common characteristic of voters’ preferences is a good approximation to the way most people think about electoral choices in the Basque region. Fourth, Nationalism as big a total effect (or even bigger) on the vote as Ideology. The reason is not because Nationalism is more salient or has greater weight in voters’ minds, but because of the alignment of the parties and the distribution of voters’ preferences in the space. Specifically, there is more agreement within the Basque electorate on Left-Right ideology than there is on the policy toward Nationalism, as reflected in the standard deviation of the preferences on the two scales. The last point is subtle. Often when survey researchers and political psychologists write about the weight of various issues in voter judgment, they conflate the true salience of the issues, the alignment of the parties across the issues, and the distribution of voters’ preferences. We have derived a straightforward way to isolate salience, and have reached a somewhat surprising conclusion. Even though voters in the Basque region put much greater weight on Left-Right ideology than they do on Nationalism, Nationalism nonetheless has a much more substantial effect in predicting the division of votes across parties. Nationalism matters more in Basque regional elections for

32 two reasons. First, the parties differentiate much more cleanly along this dimension. Along the Nationalist dimension the PNV and EH/HB/Bildu take highly nationalistic positions, locating at 8.1 and 8.7, respectively. The PP and PSE stake out similar turf at 2 and 3, respectively. Along the 10 point Left-Right scale, the parties are more widely dispersed: IU-EB/Greens locate at 2, EH/HB/Bildu at 3, PSE at 5, PNV at 6, and PP at 9. Second, the electorate hold more disparate views on Nationalism than on Left-Right ideology. Most Basque voters have a Center-Left orientation. But there is a more uniform, or even bipolar, distribution of preferences over Nationalism.

6 Conclusions

We have sought to understand why the Basque electoral system looks the way it does. The Basque Autonomous Community has a multiparty proportional representation system with an electorate divided across at least two important political dimensions (Left-Right and Nationalism) and further divided by strong cultural identities. This is a setting that seems destined to chaos. Yet, Basque regional elections produce a stable alignment of political parties and a stable pattern of minority and coalition government with the PNV almost always holding power. One of the most striking features of the Basque party system is the absence of a centrist party. There is no Center Left party that is Moderate on Nationalism, even though that is the modal voter and the median (in multiple directions). The reason for this stable party alignment, we argue, lies with the strong Identity politics tied to Basque language and culture. Euskerra speakers and others who identify with the culture vote in line with that identity, above and beyond their preferences about regional autonomy, education policy, and other policies that reflect Nationalism and quite apart from the usual Left-Right divisions common to most European political systems. The presence of such strong identity voting blunts the normally centrist pull of ideology. Consider, again, Figure 4. The presence of identity issues moves voters who are otherwise centrists toward the parties that align with their identities. This

33 has two effects. First, it pulls people away from a centrist position. Those who are normally centrist are pulled by their identity toward the Basque parties. There are fewer centrist voters than there appear to be. Second, there is more instability among the centrist voters, as some are pulled toward the Basque parties and some toward the Spanish parties. This makes it harder for a truly centrist party to establish a reliable electoral base in the presence of a strong Identity vote. Identity politics makes the center difficult for any party to hold. This is not to say that voting one’s identity rather than one’s interest or ideology is a bad thing for the individual or for the society’s politics. Identity politics in the Basque country contributes to the stability of the multi-party system: it makes it more difficult for parties to split and merge or to shift their political orientation. Identity politics prevents the sort of chaos one associates with multi-party systems where there are multiple dimensions or cleavages in the political sphere. By the same token, it may also make it more difficult to form broad coalitions that span the Euskerra-Spanish divide. Identity politics, combined with the patterns of spatial voting we have observed, contributes to dominance of nationalist parties for three decades. One of the most sur- prising findings is that Left-Right political divisions are much more salient to Basque voters than are issues concerning the degree of regional autonomy (what we have termed Nationalism). The reason, we think, is because questions of autonomy are often con- flated with individuals’ strong identities. One may vote Basque without voting for greater regional autonomy. The relatively high salience of Left-Right ideology among the electorate makes it much harder for parties arrayed along that dimension to find common ground, precisely because each party’s voters are more reluctant to compro- mise on such matters. It is very hard to bring PP and PSE under the same coalition; likewise, it is hard to bring Bildu and PNV under the same tent. Identity voting cuts in the other direction and creates a schism between the Spanish and Basque-identity parties. Voters who choose Bildu for reasons of cultural identity do not want a gov- ernment that compromises that identity through as coalition with the PSE. Perhaps

34 even more strongly felt is the division over identity between voters on the right; the PNV and PP would make highly unlikely bedfellows, precisely because of questions of identity. Against the background of strong identity and ideological voting in regional elec- tions, the politics of government formation comes into clearer focus. Identity politics can be viewed as a cost associated with cross-identity coalitions, such as Bildu-PSE or PP-PNV. Forming such cross-identity coalitions might alienate some voters who chose a party precisely because of identity. There is, then, a political price for such coalitions that is not reflected in the compromises one makes in negotiating which pub- lic policies to implement or which party holds which ministries. The political cost of crossing cultural identities appears to be quite steep in Basque politics. As a result, the most credible and likely governments are minority government by the largest party, the PNV, or coalitions among the nationalist parties.

35 References

[1] Abadie A. and Gardeazabal J. (2003) The Economic Costs of Conflict: A Case Study of the Basque Country, The American Economic Review, Vol. 93(1): 113- 132

[2] Ansolabehere, S., J. M Snyder, A. B. Strauss and M. M. Ting (2005). Voting weights and formateur advantages in the formation of coalition governments. American Journal of Political Science, 49(3): 550-563.

[3] Ansolabehere S. and J. Snyder (2000), Valence politics and equilibrium in spatial election models. Public Choice 103: 327-336.

[4] Aragon`es, E. and T. R. Palfrey (2002). Mixed equilibrium in a Downsian model with a favored candidate. Journal of Economic Theory, 103(1), 131-161.

[5] Ashforth, B. E., and F. Mael (1989). Social identity theory and the organization. Academy of management review, 20-39.

[6] Baron D. P. (1994). Electoral competition with informed and uniformed voters. American Political Science Review, 33-47.

[7] Baron D.P. and Diermeier D. (2001). Elections, governments, and parliaments in proportional representation systems. The Quarterly Journal of Economics, 116(3): 933-967.

[8] Bourne A.K. (2010) Political Parties and Terrorism: Why ban Batasuna? Paper presented at Elections, Public Opinion and Parties Annual Conference 2010.

[9] Cox, G. W. (1997) Making Votes Count. Cambridge, UK: Cambridge University Press.

[10] Diermeier D., H. Eraslan and A. Merlo (2003) A structural model of government formation. Econometrica 71(1), 27-70.

[11] D´ıezMedrano J. (1995). Divided Nations: Class, Politics, and Nationalism in the Basque Country and Catalonia. Ithaca: Cornell UP.

36 [12] Downs, A. (1957). An economic theory of democracy. New York: Harper & Row.

[13] Groseclose T. (2001). A model of candidate location when one candidate has a valence advantage. American Journal of Political Science, 862-886.

[14] Grossman, G. M., and E. Helpman (1996). Electoral competition and special interest politics. The Review of Economic Studies, 63(2), 265-286.

[15] Laver M. and Schofield N. (1998). Multiparty government: The politics of coalition in Europe. University of Michigan Press.

[16] Laver M. and Shepsle K. A. (Eds.) (1994). Cabinet ministers and parliamentary government. Cambridge: Cambridge University Press.

[17] Mart´ınez-HerreraE. (2002). From nation-building to building identification with political communities: consequences of political decentralisation in Spain, the Basque Country, Catalonia and Galicia, 1978–2001. European Journal of Political Research, 41(4): 421-453.

[18] M¨ullerand Strom (2000) Coalition Governments in Western Europe. Oxford Uni- versity Press.

[19] Myerson R. B. (1999). Theoretical comparisons of electoral systems. European Economic Review, 43(4): 671-697.

[20] Schofield N. (2003), Valence competition in the spatial stochastic model, Journal of Theoretical Politics 15: 371-383.

[21] Schofield, N. (2004). Equilibrium in the spatial ‘valence’model of politics. Journal of Theoretical Politics, 16(4): 447-481.

[22] Schofield N. and I. Sened (2005) Modeling the interaction of parties, activists and voters: Why is the political center so empty? European Journal of Political Research 44:355-390. for w

[23] Tajfel, H. (Ed.). (2010). Social identity and intergroup relations. Vol. 7. Cambridge University Press.

37 [24] Zulaika, J. (2001) Basque Violence: Metaphor and Sacrament. University of Nevada Press.

38 Table 1: Basque Parliament Election Results, 1980-2012 Political Party Percent of Votes and Number of Seats Won Year PNV- PSOE/ PP EH/HB Green Aralar EA EE UPyD/ PCE/ Other EAJ PSE-EE Bildu∗ EB-IU UA PCTV 2012 34.6% 19.1% 11.7% 25.0% 1.9% 7.6% 27 16 10 21 1 0 2009 38.6% 30.7% 14.1% 3.5% 6.0% 3.7% 2.2% 1.3%∗∗ 30 25 13 1 4 1 1 0 2005 38.7% 22.7% 17.4% 5.4% 2.3% 12.4% 1.1% 29 18 15 3 1 9 0 2001 42.7% 17.9% 23.1% 10.1% 5.6% 0.6% 33 13 19 7 3 0 1998 28.0% 17.6% 20.1% 17.9% 5.7% 8.7% 1.3% 0.3% 21 14 16 14 2 6 2 0 1994 29.8% 17.1% 14.4% 16.3% 9.2% 10.3% 2.7% 0.1% 22 12 11 11 6 8 5 0 1990 28.5% 19.9% 8.2% 18.3% 11.4% 7.8% 1.4% 0.1% 22 16 6 13 9 6 3 0 1986 23.7% 22.1% 4.9% 17.5% 15.8% 10.9% 5.2% 17 19 2 13 13 9 0 1984 42.0% 23.1% 9.4% 14.7% 8.0% 3.0% 32 19 7 11 6 0 1980 38.1% 22.7% 4.8% 16.5% 9.8% 4.0% 4.0% 25 15*** 2 11**** 6 1 0

*These parties run in coalition and under different names over time. ** Additionally, 8.8% invalidated their ballots intentionally. *** Includes votes and seats for the UCD. **** Not seated.

39 Table 2: Basque Parliament Governments, 1980-2012 Year Parties in Government Lehendakari (Party) Government 2012 PNV Urkullu (PNV) Minority Government 2009 PSOE+PP Lopez (PSE) Coalition (PNV coalition failed) 2009 (UPyD voted for Lopez) 2005 PNV-EA+EB-IU Ibarretxe (PNV) Coalition 2001 PNV-EA+EB-IU Ibarretxe (PNV) Coalition 1998 PNV+EA+EH* Ibarretxe (PNV) Coalition 1994 PNV+PSOE+EA Ardanza (PNV) Coalition 1991 PNV+ PSOE Ardanza (PNV) Coalition 1991 PNV+ EA+EE Ardanza (PNV) Coalition 1987 PNV+ PSOE Ardanza (PNV) Coalition 1985 PNV Ardanza (PNV) Minority Government 1984 PNV Garaikoetxea (PNV) Minority Government 1980 PNV Garaikoetxea (PNV) Minority Government *No Cabinet Member

* Only for 1994, 1998, 2001, and 2012, when these parties were on the ballot.

40 Table 3: Left-Right Orientations of the Basque Electorate Average Score Year (Month) Left-Right Ideology (1 to 10) of Survey Average Median SD % Missing

2012 (9) 4.04 4 1.60 16.2 2009 (3) 4.10 4 1.62 19.8 2009 (1) 4.04 4 1.46 20.1 2005 (4) 4.09 4 1.59 21.5 2005 (3) 4.25 4 1.57 19.1 2001 (5) 4.24 4 1.76 25.6 2001 (4) 4.39 5 1.83 25.2 1998 (10) 4.20 4 1.93 20.2 1994 (11) 4.19 4 1.81 15.4

* Language in 1998 is coded from 1 (Spanish exclusively spoken at home) to 5 (Euskerra exclusively spoken at home).

41 Table 4: Nationalist Orientations of the Basque Electorate Average Score Year (Month) Nationalism (1 to 10) of Survey Average Median SD % Missing

2012 (9) 4.98 5 3.06 7.6 2009 (3) 5.56 5 2.83 10.9 2009 (1) 5.04 5 2.90 10.5 2005 (4) 5.35 5 2.82 11.6 2005 (3) 5.64 5 2.63 9.6 2001 (5) 5.95 6 2.63 15.9 2001 (4) 5.70 5 2.69 12.9 1998 (10) 6.17 6 2.59 15.6 1994 (11) 6.30 6 2.39 14.0

Table 5: Left-Right and Nationalist Orientations of the Basque Electorate Nationalism (Scale: 1 = Minimal, 10 = Maximal) Ideology Min (1 to 3) Mod (4 to 6) Max (7 to 10)

Left (1 to 3) 7.9% 10.2% 18.3% Centrist (4 to 6) 12.7% 22.8% 20.1% Right (7 to 10) 3.1% 1.6% 3.3% N = 15,303

Table 6: Spanish or Basque Identities of the Basque Electorate Year (Month) Spanish or Basque Identity? Spanish Both Basque NC, NS 2012 (9) 11.3 37.4 43.9 7.5 2009 (3) 9.3 36.4 49.7 4.7 2009 (1) 11.1 37.6 47.0 4.4 2005 (4) 8.8 38.8 46.8 6.2 2005 (3) 10.4 34.7 48.4 6.5 2001 (5) 8.6 37.0 47.1 7.3 2001 (4) 9.9 37.3 46.5 6.2 Total 9.9 37.2 46.9 6.0 Sample 1,663 6,233 7,862 1,008

42 Table 7: Perceived Positions of Parties Year Party Dimension 1998 2001 2005 2009 2012 PNV Left-Right 6.2 6.5 6.3 6.6 Nationalism 8.0 8.4 8.4 8.0 7.8

PSE Left-Right 4.7 4.5 4.4 4.7 Nationalism 3.3 3.1 3.5 3.5 3.1

PP Left-Right 8.4 8.7 8.6 8.6 Nationalism 2.1 2.0 1.7 1.8 1.7

EH-B/ Left-Right 2.1 2.1 HB Nationalism 8.9 8.4 8.7

IU Left-Right 3.1 2.9 3.2 2.6 Nationalism 4.0 4.5 4.9 4.5 4.3

EA Left-Right 4.9 5.0 4.8 Nationalism 7.9 8.3 8.1 7.7 UpyD/UA Left-Right 7.6 6.6 Nationalism 2.5 2.6 2.4

43 Table 8: CIS Survey Results for Basque Parliament Election, 1994-2012 Political Party Percent of Votes and Number of Seats Won Year PNV- PSOE/ PP EH/HB Green Aralar EA UPD/ PCE / Other EAJ PSE-EE EB/EHB EB-IU UA PCTV N of Obs 2012 37.1% 18.8% 5.8% 28.9% 8.5% 0.9% 3.3% 1,471

2009 39.3% 32.0% 6.6% 4.8% 4.1% 4.3% 1.3% 7.6% 1,204

2005 46.4% 23.5% 7.6% 8.5% 3.8% 0.2% 8.9% 1.1% 2,154

2001 55.5% 14.0% 7.2% 9.6% 6.4% 7.2% 0.7% 2,708

1998 36.4% 12.9% 12.8% 22.8% 7.0% 6.1% 1.1% 1.0% 1,090

1994 52.7% 18.1% 5.2% 8.3% 6.4% 4.5% 1.8% 0.7% 1,481

44 Table 9: Orientations and Vote for Party, 1994-2012 Nationalism Ideology (Scale: 1 = Minimal, 10 = Maximal)

Percent Voting for PNV Min (1 to 3) Mod (4 to 6) Max (7 to 10) Left (1 to 3) 5.2% 20.2% 31.2% Moderate (4 to 6) 10.8% 51.1% 86.4% Right (7 to 10) 5.0% 30.7% 91.1%

Percent Voting for PSE Left (1 to 3) 64.6% 32.9% 2.0% Moderate (4 to 6) 55.0% 27.0% 2.3% Right (7 to 10) 2.5% 6.3% 0.9%

Percent Voting for PP Left (1 to 3) 1.2% 0.9% 0.0% Moderate (4 to 6) 18.9% 8.0% 0.4% Right (7 to 10) 67.5% 44.9% 1.8%

Percent Voting for EH/HB/Bildu* Left (1 to 3) 9.5% 15.2% 60.9% Moderate (4 to 6) 1.0% 3.7% 7.2% Right (7 to 10) 0.4% 1.0% 3.7%

Percent Voting for IU-EB/Green Left (1 to 3) 18.5% 29.8% 5.0% Moderate (4 to 6) 3.0% 5.6% 1.0% Right (7 to 10) 0.0% 3.9% 0.9%

Percent Voting for UPyD/UA Left (1 to 3) 0.6% 0.2% 0.0% Moderate (4 to 6) 3.8% 1.1% 0.2% Right (7 to 10) 19.5% 8.5% 0.8%

45 Table 10: Left-Right and Nationalist Orientations and Non-Voting in Basque Elections, 1994-2012 Nationalism (Scale: 1 = Minimal, 10 = Maximal) Ideology Min (1 to 3) Mod (4 to 6) Max (7 to 10)

Left (1 to 3) 41.4% 44.6% 30.0% Moderate (4 to 6) 53.9% 50.3% 26.3% Right (7 to 10) 21.2% 32.4% 13.3%

46 Table 11: Explaining Vote for Party, 2001-2012, Multinomial Logit Analysis

Independent Variable Nationalism Left-Right Economy Euskerra Population Year Choice b (SE) b (SE) b (SE) b (SE) b (SE)

2012 PSE v PNV -.57 (.04) -.79 (.09) -.02 (.12) -.98 (.24) -.06 (.08) PP v PNV -.79 (.10) 1.03 (.15) .26 (.24) -.88 (.56) .15 (.14) Bildu v PNV .17 (.04) -1.33 (.08) -.18 (.12) .88 (.20) .20 (.07) IU v PNV -.46 (.06) -1.23 (.12) -.09 (.18) -.30 (.34) -.05 (.18)

2009 PSE v PNV -.82 (.06) -.81 (.10) -.07 (.13) -.59 (.24) .19 (.08) PP v PNV -.77 (.08) 1.04 (.16) .12 (.20) -1.17 (.56) .44 (.14) IU v PNV -.48 (.08) -1.54 (.16) .40 (.20) -.85 (.40) .07 (.13)

2005 PSE v PNV -.98 (.06) -.29 (.09) .25 (.15) -.88 (.24) .10 (.12) PP v PNV -1.37 (.11) 1.56 (.19) .49 (.23) -1.55 (.54) .39 (.22) IU v PNV -.68 (.06) -.85 (.11) .32 (.18) -.24 (.26) .18 (.14) PCTV v PNV .13 (.07) -1.37 (.11) .58 (.16) .75 (.25) .19 (.12)

2001 PSE v PNV -.88 (.06) -.44 (.08) .25 (.15) .10 (.12) PP v PNV -.98 (.07) .78 (.09) .36 (.18) .75 (.16) IU v PNV -.63 (.07) -.78 (.11) .26 (.21) .56 (.21) EH v PNV .32 (.07) -.90 (.11) .46 (.16) -.35 (.12)

1998 PSE v PNV -.63 (.08) -.77 (.11) -1.45 (.20)* .08 (.10) PP v PNV -.84 (.10) .56 (.11) -1.17 (.20) .10 (.12) IU v PNV -.53 (.06) -1.24 (.13) -.50 (.22) .46 (.14)

47